Visualization, modeling and validation of chromatin interaction data

点击翻译按钮获取中文摘要

The three dimensional (3D) organization of mammalian genomes is tightly linked to gene regulation, as it can reveal the physical interactions between distal regulatory elements and their target genes. Several recent high- throughput technologies based on Chromatin Conformation Capture (3C) have emerged (such as 4C, 5C, Hi-C and ChIA-PET) and given us an unprecedented opportunity to study the higher-order genome organization. Among them, Hi-C technology is of particular interest due to its unbiased genome-wide coverage that can measure chromatin interaction intensities between any two given genomic loci. However, Hi-C data analysis and interpretation are still in the early stages. One of the main challenges is how to efficiently visualize chromatin interaction data, so that the scientific community to visualize and use it for their own research. In addition, due to the complex experimental procedure and high sequencing cost, Hi-C has only been performed in a limited number of cell/tissue types. Finally, the underlying mechanism of chromatin interactions remains largely unclear. Therefore, the PI will propose the following aims: Aim 1. Build an interactive and customizable 3D genome browser. We will build an interactive and customizable 3D browser, which allows users to navigate Hi-C data and other high-throughput chromatin organization data, including ChIA-PET and Capture Hi-C. We have built a prototype of the 3D genome browser (www.3dgenome.org). Our browser will allow users to conveniently browse chromatin interaction data with other data types (such as ChIP-Seq and RNA-Seq) from the genomic region in the same window simultaneously. Our system will also empower the users to create their own session and query their own Hi-C and other epigenomic data. Aim 2. Impute chromatin interaction using other genomic/epigenomic information. We will predict Hi-C interaction frequencies using other available genomic and epigenomic data in the same cell type, such as ChIP-Seq data for histone modifications and transcription factors. We will build our prediction model and then systematically impute Hi-C interaction matrices for all 127 cell types whose epigenomes are available thanks to recent effort by the ENCODE and Roadmap Epigenome projects. Aim 3. Perform validation experiments for computational method in aim 1 and 2. We will perform 20 3C experiments in hESC and GM cell lines, coupled with genome engineering by CRISPR/Cas9, to evaluate Hi-C prediction method in aim 2.